Does Financial Literacy Increase

Students’ Perceived Value of Schooling?

Luca Maria Pesando*

January 2017

PRC WP2017

Pension Research Council Working Paper

Pension Research Council

The Wharton School, University of Pennsylvania

3620 Locust Walk, 3000 SH-DH

Philadelphia, PA 19104-6302

Tel.: 215.898.7620 Fax: 215.573.3418

Email: prc@wharton.upenn.edu

http://www.pensionresearchcouncil.org

* Ph.D. candidate, Population Studies Center, University of Pennsylvania. Contact address: lucapes@sas.upenn.edu. The author gratefully acknowledges the financial support provided by the Fulbright Commission and the School of Arts and Sciences at the University of Pennsylvania. The author thanks Professor Olivia S. Mitchell for helpful guidance on the development of the paper, and Chiara Monticone and Angela Romagnoli for assistance with the data The author is grateful for support provided by the Pension Research Council and Boettner Center at the Wharton School of the University of Pennsylvania. All findings, interpretations, and conclusions of this paper represent the views of the authors and not those of the Wharton School or the Pension Research Council. © 2017 Pension Research Council of the Wharton School of the University of Pennsylvania. All rights reserved.

Does Financial Literacy Increase

Students’ Perceived Value of Schooling?

Luca Maria Pesando

Abstract Using data from the 2012 Programme for International Students Assessment (PISA) for Italy, this paper investigates whether financial literacy skills play a role in shaping the value that high school students place on schooling. We hypothesize that higher financial literacy may foster students’ awareness of the financial and non-financial benefits of gaining additional education, together with the costs associated with poor school outcomes. We complement OLS estimates with an instrumental variable (IV) approach to recover a plausibly causal effect of financial literacy on the school outcomes of interest, namely (a) truancy and time spent on homework outside of school (time commitment to education), and (b) attitudes towards school (attitudes). Results suggest that higher financial literacy increases students’ perceived value of schooling by boosting their time commitment to education. Conversely, there is no evidence that financial literacy shapes students’ attitudes towards school. We see this finding as consistent with the idea that adolescents’ behavior is easier to measure objectively and reliably than attitudes.

JEL Classification: A21, D10, I22, I23

Keywords: Financial literacy, schooling, truancy, attitudes, youth, Italy

I. Introduction

Financial illiteracy among the youth is widespread (Chen and Volpe 1998; Mandell 2008; Johnson

and Sherraden 2007; Lusardi, Mitchell, and Curto 2010; Lusardi 2015a), and a solid body of research has

shown that poor financial decisions on the part of students might have serious consequences for later-life

outcomes (Xiao et al. 2014; Lusardi and Tufano 2015; Brown et al. 2016). Nevertheless, very little is known

about how financial literacy at young ages interacts with school-related outcomes. In particular, no research

has focused on financial literacy as a skill shaping students’ perception of the intrinsic value of education.

This paper seeks to fill this gap.

Financial literacy is an important component of sound financial decision-making. Lusardi and

Mitchell (2014) define it as “people’s ability to process economic information and make informed decisions

about financial planning, wealth accumulation, debt, and pensions.” Yet what financial literacy is, what it

2

measures, and when it develops are still the subject of scholarly debate (Xu and Zia 2012; Fernandes, Lynch

Jr., and Netemeyer 2014). In this study we welcome the idea that financial literacy embodies a multi-

dimensional set of skills that originate early in life in the context of the family (Chiteji and Stafford 1999;

Lusardi, Mitchell, and Curto 2010; Mahdavi and Horton 2014; Grohmann, Kouwenberg, and Menkhoff

2015) and is shaped throughout the life course by means of interactions with the external context, including

the school environment.

Schooling is the most traditional form of human capital investment. As financial literacy skills entail

the ability to assess whether an investment is worthwhile or not, in this paper we explore whether more

financially literate students place a higher value on schooling as compared to less literate ones. Specifically,

do financially literate students “care” more about schooling due to their ability to discern the nature of

education as an investment which entails short-term costs and long-term benefits? We posit that students

with higher financial literacy might be more aware of the financial and non-financial (psychological and

opportunity) costs that education entails,1 as well as the monetary and non-monetary (e.g. occupational

prestige) returns that it yields. As no variable perfectly captures the value that students attribute to

schooling, in this analysis we rely on proxies such as measures of truancy and time spent doing homework

(henceforth, time commitment to education), and attitudes towards school (henceforth, attitudes). With these, we

examine whether financial literacy affects: i) the chance that students skip school days or classes; ii) the

amount of time students devote to homework outside of school; and iii) students’ attitudes towards school.

The goal of evaluating financial knowledge of high school students around the world has recently

been taken up by the OECD’s Programme for International Student Assessment (PISA), which in 2012

added a module on financial literacy to its review of high school students’ proficiency in mathematics,

reading, and science. The PISA financial literacy module has been administered to students in 18 OECD

1 As public high school education in Italy is mostly free, the opportunity cost argument is likely to prevail in this specific context. Yet we suspect individuals with higher financial literacy will be more forward-looking, hence more likely to anticipate the financial costs that college will entail.

3

and non-OECD economies that voluntarily opted to be part of the assessment.2 Around 29,000 students

completed the module, representing about nine million 15-year-olds in the schools of the 18 participating

countries and economies.3 This analysis is based on the financial literacy module in the PISA data for Italy,

one of the few participating countries that permits a regional-level analysis thanks to the inclusion of a

regional identifier and an adequate sample size.4

This paper makes three important contributions. First, we shed light on whether financial literacy

shapes individual perceptions of the inherent value of schooling among high school students, which no

prior research has heretofore addressed. In doing so, we provide a systematic description of regional

differences in financial literacy which offers a quite novel focus within the Italian context. Second, we draw

on the OECD PISA database for Italy, one of the countries with the lowest levels of financial literacy

among those that took part in the 2012 international assessment (OECD 2012a; Lusardi 2015a). Despite its

cross-sectional nature, this database has great potential and is still largely unexplored in financial literacy

research.5 Third, drawing on different data sources, we propose a set of novel and plausibly exogenous

instrumental variables to control for omitted variable bias, reverse causation, and random measurement

error in financial literacy.

Results show that higher financial literacy increases students’ perceived value of schooling by

boosting their time commitment to education. Specifically, we observe lower levels of unjustified absences

and class delays, together with more time devoted to homework in response to higher financial literacy.

Different model specifications suggest, for instance, that an increase in one proficiency level6 is associated

with one to two more hours per week spent on out-of-school study time. Conversely, there is no robust

2 Thirteen are members of the OECD: Australia, the Flemish Community of Belgium, the Czech Republic, Estonia, France, Israel, Italy, New Zealand, Poland, the Slovak Republic, Slovenia, Spain, and the United States; five are partner countries and economies: Colombia, Croatia, Latvia, the Russian Federation, and Shanghai-China (Table A1 in Appendix A). 3 Refer to the data section for an accurate description of the sampling procedure. 4 The sample is representative at the national and regional level. 5 Bottazzi and Lusardi (2016) and Christelis, Georgarakos, and Lusardi (2016) stand out as notable exceptions. 6 An increase of one proficiency level corresponds to a 75-unit increase on a 0-1000 score scale. See section 3 for further details.

4

evidence that financial literacy shapes attitudes towards school, suggesting that young adolescents’ behavior

might be easier to measure objectively than attitudes.

Our findings have wide ranging implications for educational policy. Demonstrating that financial

literacy increases students’ awareness of the importance of education is useful in informing governments

and their policy advisers as they consider new initiatives for financial education and evaluate the possibility

of including financial literacy in school curricula.

The rest of the paper is organized as follows. Section 2 discusses the available literature and

identifies the gaps the present study seeks to address. Section 3 describes the PISA financial literacy

assessment, the sample of study, and the variables chosen. Section 4 provides descriptive statistics on the

sample of schools and financial literacy skills by region. Section 5 provides empirical evidence on the

relationship between financial literacy and students’ perceived value of schooling. Section 6 summarizes the

results, discusses policy recommendations, and outlines implications for future research.

II. Literature Review

Financial literacy research has grown rapidly over the past two decades. Scholars have extensively

documented that financial literacy impacts individual decision making in several domains, with wide-ranging

implications that apply to individuals, communities, countries, and society as a whole (Lusardi and Mitchell

2014; Lusardi 2015b). Yet little is known about how financial literacy skills relate to outcomes in the teenage

years. Despite the increasing availability of data on students’ financial literacy, scant academic research has

focused on the interplay between financial literacy and school-related environments and processes.

Moreover, despite scholarly recognition that endogeneity in financial literacy is a serious concern, assessing

the validity of novel instruments exploiting regional-level variation may further contribute to enriching the

field.

a. Financial literacy as predictor of later-life behavior

5

Scholars have devoted a great deal of attention to the role of financial literacy in affecting later-life

behavior such as wealth accumulation (Jappelli and Padula 2013; Behrman et al. 2012; Fort, Manaresi, and

Trucchi 2016; van Rooij, Lusardi, and Alessie 2012), debt load (Huston 2012; Lusardi and Tufano 2015),

retirement planning (Lusardi and Mitchell 2007; Alessie, van Rooij, and Lusardi 2011; Almenberg and Säve-

Söderbergh 2011; Fornero and Monticone 2011; Lusardi and Mitchell 2011; Sekita 2011; van Rooij, Lusardi,

and Alessie 2012; Agnew, Bateman, and Thorp 2013), and stock market participation (van Rooij, Lusardi,

and Alessie 2011; Arrondel, Debbich, and Savignac 2012; Xia, Wang, and Li 2014; Arrondel, Debbich, and

Savignac 2015). In this study we explore the importance of focusing on financial literacy research for earlier-

life outcomes, endorsing the idea that financial literacy originates early in life in the context of the family

(Hastings, Madrian, and Skimmyhorn 2013; Grohmann, Kouwenberg, and Menkhoff 2015) and develops

jointly with other cognitive and non-cognitive skills, exhibiting properties such as self-productivity and

dynamic complementarity (Cunha and Heckman 2007; Cunha, Heckman, and Schennach 2010; Hastings,

Madrian, and Skimmyhorn 2013).7 If financial literacy evolves throughout the life course by means of

interactions with the external context including the school environment, this study will show whether it can

determine life outcomes well before individuals enter the labor market.

Grohmann, Kouwenberg, and Menkhoff (2015) conducted one of the first studies to systematically

assess the role of childhood experiences as root drivers of financial literacy. Using 12 childhood-related

variables, the authors explored the two main channels through which early-life experiences affect financial

literacy, namely family and schooling. They documented stronger direct influences of the former, suggesting

that the importance of childhood experiences may partially explain why it is so difficult to train and improve

young adults’ financial literacy through ad hoc courses. Their study concluded that financial literacy - and

possibly even more financial behavior - may be deeply rooted in personality traits, hence the family and

7 Self-productivity refers to the property whereby the skills produced at one stage augment the skills attained at later stages. Dynamic complementarity entails that skills produced at one stage raise the productivity of investment at subsequent stages. Together, self-productivity and dynamic complementarity produce skill multiplier effects, which are the mechanisms through which skills beget skills and abilities beget abilities (Cunha and Heckman 2007).

6

educational background of the target group should be more carefully considered when designing training

courses.

A growing body of related literature revolves around the concept of financial socialization, i.e. the

idea that children learn about finances through deliberate instruction, participation, and practice, as well as

through direct observation and involuntary interaction (Shim et al. 2010; Jorgensen and Savla 2010; Van

Campenhout 2015). Shim et al. (2010) highlighted the prominent role of the family in young adults’ financial

socialization, claiming that parents can be both direct teachers and useful role models in the financial

development of their children. Their findings suggest that seeing one’s parents as financial role models is

linked to both students’ complying in positive ways with parental expectations and to displaying favorable

attitudes towards performing healthy financial behaviors. What is more, when parents are actively involved

and serve as role models, young adults feel they can control their financial behaviors better.

b. Financial literacy in schools

Research on financial literacy in schools has taken two main directions. The first is mainly

descriptive, focusing on assessing how well-informed people are before entering the labor market.

Measuring youth’s competence in financial literacy is crucial, as in many countries young people face one of

the most important financial decisions in their lives, i.e. whether to invest in college or higher education. In

the US context, several authors have measured high school students’ financial literacy using data from the

Jump$tart Coalition for Personal Financial Literacy and the National Council on Economic Education,

providing a rather nuanced evaluation of what young people know when they enter the workforce (Mandell

2008). Studies on college student financial literacy (Chen and Volpe 2008; Cull and Whitton 2011) echoed

the findings for high school students: students are not knowledgeable about personal finance, and their low

levels of knowledge will limit their ability to make informed decisions.

A second research stream builds on experimental evidence and is concerned with quantifying the

impacts of financial education programs aimed at improving financial knowledge and awareness among

students. As one example, several U.S. states mandated financial education in high school at different points

7

in time, generating “natural experiments” utilized by Bernheim, Garrett, and Maki (2001), one of the pioneer

studies in this field. Evidence of the effectiveness of these programs in improving literacy is mixed. For

instance, while Mandell (2008) documented that Chicago students in ten classes slightly improved their

financial literacy after a financial literacy course, Mandell (2005) found no evidence of an increase in

knowledge, though a higher propensity to save in 17 to 19-year-old students who were taught financial

education in high schools. More recently, Lührmann, Serra-Garcia, and Winter (2015) examined the impact

of a short financial education program on teenagers in German high schools, which they found significantly

improved students’ interest in financial matters, their financial knowledge, and their ability to properly assess

the riskiness of assets. In a similar spirit, Becchetti, Caiazza, and Coviello (2013) used a difference-in-

difference estimation strategy to study the effect of a financial education course on financial literacy and

investment attitudes in a large sample of high school students in Italy. The course in finance reduced the

virtual demand for cash and increased the level of financial literacy and the propensity to read economic

articles, in both treated and control classes, compared to pre-treatment baseline levels. Similar programs

have been implemented in the US, Brazil, and Spain, and their mixed effectiveness discussed by Walstad,

Rebeck, and MacDonald (2010), Bruhn et al. (2013), and Hospido, Villanueva and Zamarro (2015),

respectively.

Research on financial literacy in school is, hence, quite narrow in scope. The dataset on the financial

literacy of high school students collected by the OECD as part of the 2012 Programme for International

Students Assessment therefore provides a unique opportunity to expand scholarly knowledge on the

interplay between financial literacy skills and school-related outcomes such as student performance, effort,

and motivation. No studies have yet drawn on these data to address questions of this kind.

c. Dealing with endogeneity in financial literacy research

In recent years, an increasing number of scholars has started to address the potential endogeneity of

financial literacy by means of instrumental variables (IV) or generalized method of moments (GMM)

techniques more broadly. To provide a background for what follows, we briefly review some of the studies

8

that were most successful in handling these methodological concerns. We group them according to three

broad categories of instrumental variables chosen, namely i) family background and financial knowledge of

the peer or reference group, ii) information on past education and previous financial or math knowledge,

and iii) instruments that exploit natural experiments or geographical variation in specified outcomes.

Van Rooij, Lusardi, and Alessie (2011) evaluated the relationship between financial literacy and stock

market participation instrumenting financial literacy with the financial experiences of siblings and parents,

claiming that these were not under the control of the respondent and were hence exogenous with respect to

his or her actions. Arrondel, Debbich, and Savignac (2012) attempted to answer the same question in the

French context, focusing on whether parents planned for retirement, owned a life insurance policy, or read

the economic and financial press as proxies for parental financial knowledge. Alessie, van Rooij, and Lusardi

(2011) were instead concerned with estimating the causal effect of financial knowledge on retirement

planning in the Netherlands, and they chose as IV information on whether the financial situation of the

oldest sibling was “better,” “the same,” or “worse” than the financial situation of the respondent. The same

methodology was employed by Agnew, Bateman, and Thorp (2013) to address this question in the

Australian context. Lastly, together with age-dependent variables and respondent personality traits, Behrman

et al. (2012) selected measures of family background as IVs to investigate how financial literacy affected

household wealth accumulation in Chile.

Van Rooiij, Lusardi, and Alessie (2012) relied on information on exposure to financial education

acquired in school as an instrument for financial sophistication, in order to study the relationship between

financial literacy and household net worth in the Netherlands. In particular, they focused on how much of

the respondents’ education was devoted to economics. Another example of a study relying on prior financial

knowledge is Jappelli and Padula (2013), who investigated the effect of financial literacy on wealth

accumulation and national saving throughout Europe, choosing the math literacy endowment before

entering the labor market as a potentially valid instrument for the current level of financial literacy.

9

As for the third group, Lusardi and Mitchell (2009) provided one of the most compelling studies

tackling the endogeneity of financial literacy by exploiting temporal and geographical variation in U.S. state

mandates for high school financial education. Bucher-Koenen and Lusardi (2011) instead used political

attitudes at the regional level in Germany as an instrument, arguing that free-market oriented supporters

tended to be more likely to be financially literate. Finally, Klapper, Lusardi, and Panos (2013) used the

number of public and private universities in the Russian regions and the total number of newspapers in

circulation as instruments for financial literacy, finding that financial literacy affected indicators such as

having bank accounts, using bank credit, and having spending capacity.

To the best of my knowledge, only three studies have used IVs to deal with the endogeneity of

financial literacy in the Italian context. Fornero and Monticone (2011) instrumented financial sophistication

by using cost of learning and acquiring financial knowledge and information. Specifically, they used

information on whether a household member had a degree in economics or used a computer (either at

home, at work, or elsewhere). Fort, Manaresi, and Trucchi (2016) instead investigated the causal effect of

financial literacy on financial assets exploiting confidential bank information policies as instrumental

variables. Lastly, Calcagno and Urzì Brancati (2014) selected as IV the amount of credit/debit/cashline

cards held by different households to show that financially sophisticated households hold a more balanced

portfolio, with a lower share of wealth locked up in housing assets.

There is still some disagreement among scholars on the effect size of IV estimates as compared to

simple Ordinary Least Squares (OLS). Lusardi and Mitchell (2014) reported that out of 11 studies using

instrumental variables for financial literacy across different countries, all of them documented larger IV

estimates. Behrman et al. (2012), Fort, Manaresi, and Trucchi (2016), and Calcagno and Urzì Brancati (2014)

aligned with Lusardi and Mitchell’s findings (2014). Conversely, in a meta-analysis of the relationship

between financial literacy and financial behavior, Fernandes, Lynch Jr., and Netemeyer (2014) reached the

opposite conclusion. According to their study, analyses using instrumental variables produced smaller effects

than those using simple cross-sectional designs and OLS.

10

Despite the existence of papers using instrumental variable techniques to account for the

endogeneity of financial literacy, all of them have been concerned with predicting forms of financial

behavior such as stock market participation, retirement planning, or household wealth accumulation. By

contrast, this paper contributes to the literature by providing a new strategy to identify the causal effect of

financial literacy on yet unexplored school outcomes such as measures of time commitment to education

and attitudes towards school. The analysis below also sheds additional light on the IV-OLS effect size

controversy outlined above.

III. Data and Methodology

a. Sample

In what follows we use data collected by the OECD Programme for International Student

Assessment (PISA) for high school students in Italy in 2012. The goal of the PISA financial literacy module

was to assess the literacy and financial capabilities of 15 year-old students across 18 OECD and non-OECD

countries.8 This survey evaluated high school students’ competencies in reading comprehension,

mathematics, science and, for the first time in 2012, financial literacy. The latter questions were included in

response to the last major economic crisis, which evidenced a widespread lack of financial sophistication

that contributed to ill-informed financial decisions.

PISA samples were designed in a two-stage stratified fashion. First, individual schools in which 15-

year-old students could be enrolled were sampled, with probabilities proportional to their size.9 In the

second stage, students were sampled within schools.10 In the 18 countries that also implemented the

optional financial literacy assessment, sampled schools selected with equal probability 43 students, 35 of

whom completed the core assessment, and the remaining eight were administered the financial literacy

8 Findings about these data were officially released on July 9, 2014. 9 A minimum response rate of 85 per cent is required for the schools initially selected. 10 PISA 2012 also requires a minimum participation rate of 80 per cent of students within participating schools, to be met at the national level.

11

module.11 PISA has an age-based definition for its target population, i.e. one not tied to the institutional

structures of national education systems (OECD 2014a). PISA assesses students who are age between 15

years and three months and 16 years and two months at the beginning of the assessment period (plus or

minus a one-month allowable variation), and who are enrolled in an educational institution in grade seven or

higher, regardless of the grade levels or type of institution in which they are enrolled, and regardless of

whether they are in full-time or part-time education. Hence, PISA tests the knowledge of a group of

individuals born within a comparable reference period who may have followed different educational

trajectories both in and out of schools.

The 2012 PISA assessment consisted of 40 math and reading questions, as well as questions about

students’ experiences with money matters. Additional questions were asked to students to gather

information about themselves, their home and school environment, their learning experiences, and their

attitudes. School principals also answered questionnaires on school policies, the learning environment, and

the school’s provision of financial education in prior and the current year.12 In three of the 18 countries –

namely Belgium, Croatia, and Italy – parents filled a questionnaire too.

Of the 18 countries surveyed, Italy provides an interesting case study for several reasons. First,

results from the 2012 PISA financial literacy assessment show that Italy occupies the second-lowest rank out

of the 18 OECD and non-OECD countries that participated in the module (Table A1, Appendix A), with a

mean financial literacy score of 466, well below the OECD average of 500 (OECD 2014a). Second, Italy is

the country with the highest number of schools sampled (1,158) and the highest number of students within

schools (7,068), providing a substantial sample size for analysis. Third, and related to the above, the dataset

11 For additional details please refer to: OECD (2014b), PISA 2012 Technical Report, OECD Publishing, Paris. 12 Although these might be potentially good sources of exogenous variation in financial literacy, they suffer from several limitations. First, even if the school took part in a financial education program, there is no way to assess whether the single student benefited from it. Second, questions on the availability of financial education and questions on financial literacy use different “reference points” (students in the school vs students in the national modal grade for 15 year olds), making it tricky to relate them. Third, there are serious inconsistencies in the answers, likely reflecting a poor understanding of the questions. To address these shortcomings, the OECD team is in the process of developing higher-quality financial education questions.

12

for Italy includes a regional identifier that makes it possible to conduct the analysis at the sub-national level

with a satisfactory sample size per region.

b. Measures

Table 1 summarizes the school outcomes chosen for this analysis. No variable perfectly captures the

extent to which students value schooling. Nevertheless, the PISA database gives the opportunity to examine

proxy indicators such as measures of truancy, time spent on out-of-school study time, and attitudes towards

school. We presume that students who skip classes or school days frequently, think school is a waste of

time, and devote little time to homework, would be more likely to perceive schooling as a worthless

investment.13 Our analysis investigates the relationship between financial literacy and each outcome

separately. We attempted to combine truancy measures and attitudes into factors, yet any factor analysis

suggests a degree of reliability (Cronbach’s alpha) less than 0.7 - the conventionally accepted threshold for

factors. We take this evidence as suggestive that each selected variable captures a different underlying

construct.14

[Table 1 about here]

Financial literacy was measured in PISA using a mixture of multiple-choice and constructed-

response questions. These items assessed familiarity with financial products, understanding and use of

financial terms, conceptual understanding of numeracy, application of numeracy skills, and capacity to make

effective financial decisions. PISA assessed not only whether students can reproduce knowledge, but also

whether they can extrapolate from what they have learned and apply their knowledge in new contexts. Three

dimensions were considered when designing the financial literacy questions, namely content, i.e. the areas of

knowledge that are essential for financial literacy; processes, i.e. the approaches and mental strategies called

upon to negotiate the material; and contexts, i.e. the situations in which the financial knowledge, skills, and

understanding are applied (OECD 2014a).

13 Holding other factors constant, such as parental influences. 14 Simple correlations (Table A2 in Appendix A) align with this idea.

13

The PISA test design permits to construct a single scale of proficiency, drawing on all the questions

in the financial literacy assessment.15 Each question is associated with a particular point on the scale

indicating its difficulty, and each student’s performance is associated with a particular point on the same

scale that indicates his or her estimated financial literacy proficiency (OECD 2014a). The single continuous

scale is divided into five levels (Table 2). Level 1 indicates low proficiency, Level 2 is the international

baseline proficiency level, and Level 5 indicates high proficiency. Students at Level 1 are considered to be

financially illiterate, while students at Level 2 are only starting to apply their knowledge to make financial

decisions in contexts that are immediately relevant to them.16 Each proficiency level represents 75 score

points, i.e. there are 75 points between the top of one level and the top of the next.17 A difference in

performance of one proficiency level therefore represents a significant gap in financial literacy performance

(OECD 2014a). Proficiency levels allow researchers to investigate differences in financial literacy not only

across countries but also within countries (Lusardi 2015a).

[Table 2 about here]

The structure of PISA prevents the use of a single value as a proxy for the student’s results, since

each student only replies to a certain number of questions in the entire questionnaire. These replies, together

with information on several variables in the questionnaire, yield a distribution of values to be created a

posteriori for each individual. In total, five random values called plausible values (PVs) are obtained from this

distribution for each student.18 The five plausible values need to be accounted for in the estimation process

in order to avoid problems associated with biases and inefficiency (OECD 2014b). To account for these

properties, the PISA database provides 80 replicates of individual weightings, which allow for efficient

15 The OECD makes the single item responses publicly available but not the text of the questions. 16 Students at each level are expected to be proficient at the preceding level, and to reach a particular proficiency level a student must correctly answer a majority of items at that level. 17 In the analysis that follows I divide the financial literacy score by 75 to interpret a one-unit increase in financial literacy as an increase in one proficiency level. 18 As it is common with item response scaling models, student proficiencies are not observed; they are missing data that must be inferred from the observed item responses. PISA uses the imputation methodology usually referred to as plausible values (PVs) to make this inference and produce consistent estimators of population parameters. PVs are imputed values that resemble individual test scores and have approximately the same distribution as the latent trait being measured. PVs represent random draws from an empirically derived distribution of proficiency values that are conditional on the observed values of the assessment items, plus the background variables.

14

estimators. The use of replicates is necessary because of the way in which individuals are selected for the

PISA sample.19

c. Methodological approach

As the PISA study is cross-sectional, the relationship between financial literacy and the outcomes of

interest may suffer from reverse causation. In addition, these variables might be jointly determined by a

third omitted factor such as unobserved ability, personality trait, or family upbringing. Empirical measures

of financial literacy are also likely to suffer from measurement error that, ceteris paribus, can bias standard

estimates of the impacts of financial literacy towards zero. In order to deal with these concerns, this analysis

complements ordinary least squares (OLS) estimates with an instrumental variable approach (IV). We first

estimate the following OLS specification:

! = #$%$ + '(%) + *$

where ! is any school outcome proxying for students’ perceived value of schooling, #$ is a vector of

individual, household, and school-level covariates, and %) is our coefficient of interest.

For each outcome we estimate four specifications to explore how additional controls affect the

relationship between financial literacy and the dependent variable. Model 1 provides the bivariate correlation

between financial literacy and the outcome. Model 2 adds individual-level characteristics such as gender, age,

and grade attended. In Model 3 we add household-level controls such as family socio-economic status and a

dummy for whether the student lives in a household with both parents. Socio-economic status is estimated

by the PISA index of economic, social, and cultural status (ESCS), built to be internationally comparable

and based on indicators such as parents’ education and occupation, the number and type of home

possessions - which are used to indicate levels of household wealth, - and the educational resources available

at home. Lastly, model 4 accounts for school-level characteristics such as the type of school attended (public

vs private), school orientation (generic vs vocational), school size as measured by total school enrolment,

19 Stata packages have been developed to handle the PISA two-stage stratified sample design, the use of replicates, and the plausible values for the test scores. Repest and pisatools are some examples.

15

class size, and student-teacher ratio. All specifications account for regional dummies to control for region-

specific heterogeneity.20 Standard errors are clustered at the school level.

Building on the previous specification, we then attempt to account for the potential endogeneity of

financial literacy by estimating the following model:

! = #$%$ + '(%) + *$

'( = #$,$ + #),) + -)

where #$ is a vector of controls - as before - and #) is a vector of financial literacy instruments. The

candidate instrumental variables should predict '( well but they should not affect ! directly or indirectly

through other unobserved factors. Arguably, properly chosen instruments are similar to quasi experiments

when the instrument for financial literacy is not plausibly caused by the dependent variable (Angrist and

Krueger 2001).

We propose a series of regional and individual-level instruments separately, and evaluate whether a

combination of them further improves the precision of the estimates. We construct regional-level

instruments by complementing the PISA file with additional data sources. From the statistical database of

the Bank of Italy and the Census of Banks (SIOTEC), we obtain data on the number of Italian banks,

branches, and ATMs (per 1,000 inhabitants) per region in the years 2010, 2011 and 2012. From the

Chamber of Commerce and the Italian Ministry of Justice, we obtain data on Italian youth enterprises and

bankruptcies in 2012. From the Italian National Institute of Statistics (ISTAT), we gather information on

the demographic composition of high schools in each region. From Accertamenti Diffusione Stampa

(ADS), we obtain provincial-level data on newspaper circulation by region.21 Lastly, we merge historic data

constructed by Bertocchi and Bozzano (2015) on the percentage of provinces per region whose main cities

lay on medieval commercial routes or hosted a fair or a bank in the 13th and 14th centuries. This latter

variable is included to test for the role of historical persistence. i.e. to explore the hypothesis that financial

20 Regional dummies are omitted – by construction – whenever we use regional-level instruments. 21 PISA does not include province identifiers nor it is representative at the provincial level. Therefore any provincial-level data is aggregated at the regional level.

16

literacy today is higher in regions where more provinces were involved in commercial exchanges in the past

(Bottazzi and Lusardi 2016).

Individual-level instruments are devised using PISA data aggregated at the school level, excluding the

information pertaining to the student him/herself.22 We construct the percentage of a student’s schoolmates

who have at least one parent working in direct contact with money flows, the average educational

expenditure among a student’s schoolmates, the percentage of schoolmates who have a bank account or a

prepaid debit-card, and the percentage of schoolmates who discuss money matters with their parents at least

once per week.

We acknowledge that some of these variables could still affect the dependent variable directly, so we

use Hansen’s J test of over-identifying restrictions to determine which instruments appear truly independent

of the second-stage disturbance term. We conclude that the Hansen J statistic does have power in

identifying problematic candidate instruments and use only the variables that survive tests for both

instrument strength (i.e. relevance) and exogeneity (i.e. validity). Specifically, we carry out the empirical

analysis relying on a combination of three regional-level instruments (Table 3) – the share of circulating

newspapers with finance-related content in 2012,23 the growth of ATM branches between 2010 and 2012,

and the number of bankruptcies in 2012 – and one individual-level instrument – the percentage of a

student’s schoolmates with at least one parent working in direct contact with money flows.24

[Table 3 about here]

The validity of the regional-level instruments rests on the assumption that these variables should

correlate well with students’ financial literacy without affecting the school outcomes directly or indirectly

through unobserved factors. Our variable on the share of finance-related newspapers builds on the volume

of released newspapers – instead of the frequency or number of newspapers actually read – so we view this

22 No class-level identifier is included in PISA. 23 This is constructed as the volume of released newspapers with finance-related content (namely ilSole24ore newspaper) over the volume of all newspapers released in 2012. 24 This variable is built by the author using the International Standard Classification of Occupations 2008 (ISCO-08) developed by the International Labour Organization (ILO). http://www.ilo.org/public/english/bureau/stat/isco/

17

variable as less driven by demand-side factors that correlate with household-level socio-economic

considerations and more likely to exogenously capture exposure to financial information and/or to peers

(either other family members, schoolmates, or neighbors) with higher financial knowledge. Several other

studies have shown that individuals learn about financial matters from peers (Duflo and Saez 2003; Hong,

Kubik, and Stein 2004; Brown et al. 2008). By similar reasoning, we expect the number of bankruptcies in

2012 to raise students’ financial awareness by means of heightened media coverage, and the growth of ATM

branches to increase the chances that students use prepaid debit cards and check on their bank accounts and

savings regularly.25

As for the validity of the individual-level instrument, we draw on prior literature (van Rooij, Lusardi,

and Alessie 2011; Klapper, Lusardi, and Panos 2013) to argue that variables related to family background,

financial knowledge, and financial experiences of other individuals in one’s reference group are not under

the direct control of the respondent and hence are likely to be exogenous with respect to pupil behavior. We

hence posit that a higher share of schoolmates with parents working in direct contact with money may

correlate well with respondents’ acquisition of financial literacy by way of peer influence and affect students’

perception of schooling only through better financial literacy. To minimize the concern that this instrument

retains a strong SES component (e.g. parents working in finance concentrate in high-SES groups), the

variable is built such that we account for all types of occupations that require recurrent money exchanges,

i.e. from bank managers to grocery store cashiers.26

IV. Descriptive statistics

a. Dependent variables

25 F-tests for first stage are reported in Table A3 (Appendix). 26 The percentage of the respondent’s schoolmates who have a bank account and the percentage of the respondent’s schoolmates who discuss with parents about money matters at least once per week do satisfy the relevance requirement (F=35.59 and F=13.16, respectively). Yet the first question was asked only to half of the students, and the second to the other half. When combined with the fact that some of the outcomes were also not asked to some groups of students, the sample size loss is significant. Anyway, findings using these variables as instruments are unchanged - and available upon request.

18

Table 4 reports descriptive statistics on the school outcomes of interest. Despite their inherent

nature as categorical variables, we take truancy and attitudes towards school as continuous variables, with

values ranging from 0 to 3, defined as in Table 1. Questions pertaining to time spent on activities and

attitudes towards school were only asked to two-thirds of the students, so the table has fewer observations

for some of the outcomes.27 Variables on attitudes have been renamed and recoded consistently to ensure

that higher ordinal values are associated with more positive attitudes towards school. Descriptive statistics

reveal that Italian high school students skip a whole school day on average one day every two weeks, and

they get to school late or skip classes at a similar – yet somewhat lower – frequency.28 Moreover, students

spend a little less than nine hours per week on homework and, despite considerable variability, the majority

of them agree with the proposition that schooling is a worthwhile investment.

[Table 4 about here]

b. School characteristics

Table 5 describes the nature and characteristics of the 1,158 schools sampled for the PISA financial

literacy assessment, by geographic region. Around 50 schools per region were sampled, with an average of

350 students.29 Although the OECD planned to administer eight financial literacy questionnaires per

sampled school, our table shows that on average six to seven students were actually interviewed. The

percentage of private schools is below 10 per cent in almost every region, reflecting the reality that high

school education in Italy is predominantly provided by the public system. This holds even more strongly in

the South of Italy and the Islands, where the percentage of private schools drops below five per cent. In

Italy, school quality does not vary dramatically by family income level or location, as it does in some

27 The OECD devised three types of questionnaires, i.e. A, B, and C. Questions on time spent on activities were not asked in questionnaire B, while questions on attitudes towards school were not asked in questionnaire A. Even though these are not technically missing data (they are coded as N/A, i.e. not available), in the Appendix (Table A4) we verify that these “missing” students do not systematically differ in terms of socio-demographic characteristics. We find no systematic differences in terms of age, gender, grade, and socio-economic status. Yet we document statistically significant differences in the financial literacy score. Students missing on the dependent variable tend to have on average a lower financial literacy score, a finding to keep in mind when interpreting results. 28 Due to the way the variable is coded, this is an approximate assessment. 29 Trentino-Alto Adige has a far higher number of schools and students as I combined the autonomous provinces of Trento and Bolzano – separate in the PISA data - to reflect the reality that Italy has actually 20 regions (instead of 21).

19

countries such as the United States (Bottazzi and Lusardi 2016). Public schools are not considered to be

lower quality than private schools, and the difference between students attending the two types of schools

rests mainly on the socioeconomic status of the family, a variable we are able to control for. Lastly, there is

little cross-regional variability in terms of class size and student-teacher ratio, which average 25-26 students

per class and 10 students per teacher, respectively.

[Table 5 about here]

c. Financial literacy

Italy’s performance in financial literacy is second to last when considering all countries and

economies participating in the global financial literacy assessment (Table A1, Appendix A). More than one

in five students does not reach the baseline level of proficiency in financial literacy, and only two per cent of

students are top performers (Level 5). Regional differences are large (Figure 1), as the difference between

the best-performing region (Veneto) and the worst-performing one (Calabria) is around 86 score points, i.e.

larger than one proficiency level. To locate these findings in a comparative perspective, only two regions

report financial literacy scores that are above the average of the OECD countries (500), namely Veneto and

Friuli-Venezia Giulia, in the North-East part of Italy. Students in Trentino-Alto Adige and Lombardia are

close to the OECD average, yet students in the remaining regions fall far behind. Results show a clear

regional patterning of financial knowledge whereby students in the North-East and North-West score above

the country-level average (494 and 486, respectively), while students in the South and the Islands score

considerably below (443 and 432, respectively). Students attending schools in the Center of Italy score

around the average of 466 points. The data hence provide evidence of widespread financial illiteracy

throughout the country, with marked regional differences that place students in some regions of the South,

such as Calabria, 200 points – almost three proficiency levels – below their 15-year-old Chinese peers

(OECD 2014a).

[Figures 1 and 2 about here]

20

Figure 2 provides a breakdown of the share of Italian students who fall within each financial literacy

proficiency level, by region. Two findings are particularly striking. First, more than 20 per cent of students

perform below the baseline level of proficiency (Level 2) on average, with proportions ranging from eight

per cent in Friuli-Venezia Giulia to around 39 per cent in Calabria. This high share of low-performers is

alarming since some of these students cannot even complete some of the simplest financial tasks such as

recognizing the difference between needs and wants or comparing the value of goods based on a

comparison of their price per unit (OECD 2014a). On the other side of the spectrum, the share of students

reaching Level 5 or above averages below two per cent, and it surpasses four per cent in only two North-

Eastern regions, namely Trentino Alto-Adige (4.8) and Veneto (4.9).

V. Multivariate results

Table 6 provides results from an OLS specification of the school outcomes of interest on financial

literacy, accounting for individual (model 2), household (model 3), and school-level (model 4) controls

progressively. Panel (a) focuses on pupil time commitment to education outcomes, while panel (b) focuses on

attitudes. Our model specifications suggest that financial literacy is positively associated with the value

students place on schooling, regardless of whether the latter is proxied by measures of time commitment to

education or school attitudes. Coefficients are strongly statistically significant and robust to the inclusion of

individual, household, and school-level controls. Accounting for controls marginally affects the significance

of the estimates, though it does reduce the magnitude of the financial literacy coefficients by a fifth to half.

[Table 6 about here]

Among individual-level controls, gender plays a relevant role, suggesting that girls place on average a

higher value on schooling as compared to boys. Girls tend to be late for school less frequently, skip fewer

school days or classes, spend almost three additional hours per week on homework, and agree that

schooling is a beneficial investment that will help them be more prepared for college and adult life more

broadly. Age turns out to be a weak predictor of all school outcomes, which is not surprising given the little

21

variability in the age of the students sampled. We also see that students in higher grades are less likely to

engage in truancy and more likely to spend time on homework, though their attitudes differ only slightly

from those of lower-grade students.

Evidence on the role of household-level characteristics is more mixed. Except for a few outcomes,

there is no evidence of a clear SES-gradient in students’ perceived value of schooling, while family stability

and parental involvement – as proxied by an indicator for whether the student lives in a family with two

parents – emerge as clearer drivers of students’ time commitment to education. In terms of school

characteristics, students in private schools are less likely to skip school days, consistent with the nature of

the private school system which tends to devote more resources to monitoring student absences and

interacting with parents. Lastly, vocational education is associated with higher truancy and more negative

attitudes towards school, in line with the idea that students who sort into vocational education value

working more than schooling and report a higher discount rate.

The positive association between financial literacy and students’ perceived value of schooling is

consistent across outcomes. The only exception to this trend is in the “get a job” question, which suggests

that more financially literate students tend to disagree with the idea that trying hard at school will help them

get a good job.30 We suspect students might have interpreted the idea of getting a job as a short-term

substitute for college education, instead of a life stage that can occur after college and build upon prior

human capital investments. This hypothesis is consistent with both the positive financial literacy coefficient

observed in the “prepare for college” question, and the positive association between vocational education

and “get a job”, which departs from the overall trends discussed above. By suggesting that students in

vocational education agree with the claim that trying hard at school will help them get a good job – though

not get into a good college – we have yet another indication that students may not have understood the

meaning of the question. We remain confident that this exception does not invalidate our main conclusions.

[Table 7 about here]

30 In the full specification (model 4), the financial literacy coefficient is only significant at the 10 per cent level.

22

Table 7 assesses the robustness of the findings reported in Table 6 by addressing potential

endogeneity concerns. We report the financial literacy coefficient from a full specification accounting for

individual, household, and school-level controls.31 Column (1) reports the financial literacy coefficient from

an OLS specification, as per model 4 in Table 6. Columns (2), (3) and (4) report the same coefficient from

an IV specification where financial literacy is instrumented through regional-level, individual-level, and a

combination of both regional and individual-level variables. Results point to three main findings. First, there

is evidence that financial literacy plays a role in shaping students’ perceived value of schooling, yet this effect

persists only when relying on measures of time commitment to education. Second, IV estimates are aligned

in terms of sign, magnitude, and statistical significance, regardless of the type of instrumental variable used.

Third, even the most conservative IV estimate is three to four times larger in magnitude as compared to its

OLS counterpart, suggesting that OLS results underestimate the magnitude and significance of financial

literacy as a determinant of school-related outcomes.32

Estimates of the relationship between financial literacy and students’ time commitment to education

indicate that an increase of one proficiency level in financial literacy reduces the number of times students

are late for school and skip school days by 0.18 and 0.16, respectively (column 3), a close-to-fourfold

reduction as compared to the OLS estimates.33 As students get to school late and skip a whole school day

roughly one day every two weeks, this effect size implies a reduction in truancy by 18 and 16.4 per cent per

two-week period, respectively. Moreover, an increase in one proficiency level is associated with one to two

more hours per week spent on out-of-school study time, a small yet non-negligible gain. Scholars have

shown that time spent on homework has positive impacts on the educational trajectories of children as they

31 According to Table 6, this is the most conservative specification. The full output with four specifications per outcome can be found in Table A5 (Appendix A). 32 The attenuation bias due to measurement error in financial literacy also drives estimates towards zero. Measurement error in financial literacy may be substantial because respondents may guess the answer at random or they may misunderstand the question. 33 We are aware of the complexity of interpreting changes on a 0 to 3 continuous scale where each level does not correspond to an exact number of times a student skipped school or classes (Table 1), yet this parameterization makes the estimation strategy and the application of the sampling design through in-built Stata commands more straightforward. Substituting with dichotomous variables measuring the probability of being late for school or skipping one whole day at least once over the previous two weeks, the interpretation of these effects suggest that an increase in one FL proficiency level is associated with a reduction in the probability of being late for school by 11.4 per cent and a reduction in the probability of skipping one whole day by 12 per cent.

23

grow older (Eren and Henderson 2008; 2011). Specifically, while the association between time on

homework and achievement gains – as measured by grades or standardized test scores – tends to be weak

for children in elementary school, it is strong and significant for high school students (Cooper, Robinson,

and Patall 2006).34 Drawing in turn on the large body of research relating academic achievement and

subsequent wages (Currie and Thomas 1999; French et al. 2015), we suspect high school financial literacy

might have powerful indirect effects on individual wages later in life.

By contrast, the relationship between financial literacy and attitudes towards school fades once we

control for endogeneity, providing no evidence of a direct causal effect of the former. The effect size does

not statistically differ from zero across all outcomes and all specifications, hence increasing confidence in

the validity of the chosen instruments.

VI. Discussion

In this study we have used Italian data from the 2012 OECD PISA financial literacy assessment to

investigate whether financial literacy skills play a role in shaping the value that high school students place on

schooling. Our analysis posits that financial literacy may play a role within the school context by raising

students’ awareness of the financial and non-financial benefits of gaining additional education, together with

the lifetime costs associated with poor school outcomes. To measure the extent to which students value

schooling, we use measures of time commitment to education and attitudes towards school. Our results

suggest that higher financial literacy increases students’ perceived value of schooling by boosting their time

commitment to education. Yet we find no robust evidence that financial literacy shapes students’ attitudes

towards school.

The finding that financial literacy shapes students’ time commitment to education but not attitudes

towards school is at odds with the belief that changes in attitudes are a precondition for changes in behavior

34 This correlation holds up to a certain hour threshold. Cooper, Robinson, and Patall (2006) claimed that for high school students achievement continued to improve with more homework until assignments lasted around 2 and 2 and a half hours a night.

24

(Ajzen 1991; Armitage and Conner 2001). We advance three hypotheses that can help reconcile this puzzle.

First, attitude-related questions might not be a valid proxy for students’ perceived value of schooling, as

somewhat evidenced by the “get a job” question discussed above. Second, even if these measures are

reliable, we suspect adolescents’ behavior could be easier to measure objectively than attitudes, especially

during ages in which attitudes are quite unstable and the value of a higher education is not clear yet. Third,

the nature of financial literacy as a skill that stresses the practical side of decision-making may be more

conducive to immediate behavioral changes, such as reduced delays or absences, as compared to shifts in

attitudes.

In terms of time commitment to education, we document a larger effect size using IV versus OLS.

Our most conservative IV specifications indicate that an increase in one financial literacy proficiency level

implies a reduction in school delays and school absences by 18 and 16.4 per cent, respectively, over a two-

week period. Also, an increase in one proficiency level is associated with one to two additional hours per

week spent on out-of-school study time. When converted to annual terms,35 a one to two hour per week

increase in time devoted to homework translates into 34 to 68 more hours per year, a substantial human

capital advantage. This study further adds to the debate over the relative size of OLS vs IV estimates in

financial literacy studies. In this respect, our finding that OLS coefficients on financial literacy are biased

downward due to omitted variable bias and measurement error in financial literacy aligns with Behrman et

al. (2012) and Lusardi and Mitchell (2014).

More broadly, this research highlights the role of financial literacy as a driver of human capital

accumulation. By departing from math and reading skills in its deeper focus on decision-making, the value

of time, the role of incentives and rewards, and the importance of savings,36 we show that financial literacy

35 A school year has approximately 34 weeks. 36 Math and reading skills are indeed correlated with financial literacy, yet it is important to highlight that financial literacy retains a different character, to the extent that a high level of competence in the former two disciplines does not necessarily translate into a high level of financial literacy. Also, Italy is one of the countries in which variation in financial literacy is least explained by variation in math skills (OECD 2014a). In Italy about 62% (compared to the OECD average of 75%) of the financial literacy score reflects some skills also measured in math or reading assessments, while 38% (compared to the OECD average of 25%) of the score indicates factors measured only from the financial literacy assessment.

25

affects the extent to which students view schooling as a worthwhile investment by curtailing wasted human

capital. Although the data do not allow us to further investigate the precise mechanisms whereby this effect

is produced, we suspect that financial literacy shapes students’ perception of the intrinsic value of education

by raising their awareness of the short-term financial and non-financial costs that education entails, as well

as the long-term monetary and non-monetary returns that it yields. This finding is of particular relevance to

academics and policy-makers concerned with boosting human capital accumulation during the teenage years.

Additionally, one might expect positive spillover effects on other later-life outcomes such as the choice of

occupation, wealth, earnings, savings, stock-market participation, and retirement preparedness.

Policy implications are clear. In the case of Italian youth, the highest priority is in the Southern

regions. Devising effective strategies to boost financial knowledge of students in the South would allow the

national average to get closer to the OECD one. On a broader scale, governments can strengthen financial

education interventions by complementing students’ academic training. Including a course with economic or

finance-related content in school curricula could be an effective strategy, exposing students to structured

sequences providing them with the tools needed to make better financial decisions.

26

References

Agnew, J. R., H. Bateman, and S. Thorp. 2013. Financial literacy and retirement planning in Australia. Numeracy 6 (2): 1-25.

Ajzen, I. 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes 50: 179-211.

Alessie, J., M. van Rooij, and A. Lusardi. 2011. Financial literacy, retirement preparation and pension expectations in the Netherlands. NBER Working Paper No. 17109.

Almenberg, J., and J. Säve-Söderbergh. 2011. Financial literacy and retirement planning in Sweden. Journal of Pension Economics and Finance 10 (4): 585-598.

Angrist J., and A. Krueger. 2001. Instrumental variables and the search for identification: From supply and demand to natural experiments. Journal of Economic Perspectives 15 (4): 69-85.

Armitage, C., and M. Conner. 2001. Efficacy of the theory of planned behaviour: A meta-analytic review. British Journal of Social Psychology 40: 471-499.

Arrondel, L., M. Debbich, and F. Savignac. 2012. Stockholding and financial literacy in the French population. International Journal of Social Sciences and Humanity Studies 4 (2): 285-294.

Arrondel, L., M. Debbich, and F. Savignac. 2015. Stockholding in France: The role of financial literacy and information. Applied Economics Letters 22 (16): 1315-1319.

Becchetti, L., S. Caiazza, and D. Coviello. 2013. Financial education and investment attitudes in high schools: Evidence from a randomized experiment. Applied Financial Economics 23 (10): 817-836.

Behrman, J. R., O. S. Mitchell, C. K. Soo, and D. Bravo. 2012. How financial literacy affects household wealth accumulation. American Economic Review 102 (3): 300-304.�

Bernheim, B. D., D. M. Garrett, and D. M. Maki. 2001. Education and saving: The long-term effects of high school financial curriculum mandates. Journal of Public Economics 80 (3): 435-465.

Bertocchi, G., and M. Bozzano. 2016. Women, medieval commerce, and the education gender gap. Journal of Comparative Economics 44 (3): 496-521.

Bottazzi, L., and A. Lusardi. 2016. Gender differences in financial literacy: Evidence from PISA data in Italy. Global Financial Literacy Excellence Center (GFLEC) Working Paper, Washington DC.

Brown M., J. Grigsby, W. van der Klaauw, J. Wen, and B. Zafar. 2016. Financial education and the debt behavior of the young. Review of Financial Studies 29 (9): 2490-2522.

Brown, J., Z. Ivkovic, P. Smith, and S. Weisbenner. 2008. Neighbors matter: Causal community effects and stock market participation. Journal of Finance 63 (3): 1509– 1531.

Bruhn, M., L. de Souza Leao, A. Legovini, R. Marchetti, and B. Zia. 2013. The impact of high school financial education: Experimental evidence from Brazil. The World Bank, Policy Research Working Paper Series: 6723.

27

Bucher-Koenen, T., and A. Lusardi. 2011. Financial literacy and retirement planning in Germany. Journal of Pension Economics and Finance 10 (4): 565–584.

Calcagno, R., and M. Urzì-Brancati. 2014. Do more financially literate households invest less in housing? Evidence from Italy. Economics Bulletin 34 (1): 430-445.

Chen, H., and R. P. Volpe. 1998. An analysis of personal financial literacy among college students. Financial Services Review 7 (2): 107–128. Chiteji, N. S., and F. P. Stafford. 1999. Portfolio choices of parents and their children as young adults: Asset accumulation by African-American families. American Economic Review 89 (2): 377–380.

Christelis, D., D. Georgarakos, and A. Lusardi. 2016. The impact of bank account ownership on adolescents’ financial literacy. Global Financial Literacy Excellence Center (GFLEC) Working Paper, Washington, DC.

Cooper, H., J. C. Robinson, and E. A. Patall. 2006. Does homework improve academic achievement? A synthesis of research, 1987-2003. Review of Educational Research 76 (1): 1-62.

Cull, M., and D. Whitton. 2011. University students’ financial literacy levels: Obstacles and aids. The Economic and Labour Relations Review 22 (1): 99-114.

Cunha F., and J. Heckman. 2007. The technology of skill formation. American Economic Review 97 (2): 31–47.

Cunha F., J. Heckman, and S. Schennach. 2010. Estimating the technology of cognitive and noncognitive skill formation. Econometrica, 78 (3): 883–931.

Currie, J. and D. Thomas. 1999. Early test scores, socioeconomic status and future outcomes. National Bureau of Economic Research (NBER) Working Paper 6943.

Duflo, E., and E. Saez. 2003. The role of information and social interactions in retirement plan decisions: Evidence from a randomized experiment. Quarterly Journal of Economics 118 (3): 815–842.

Eren, O., and D. J. Henderson. 2008. The impact of homework on student achievement. Econometrics Journal 11 (2): 326-348.

Eren, O., and D. J. Henderson. 2011. Are we wasting our children’s time by giving them more homework? Economics of Education Review 30 (5): 950-961.

Fernandes, D., J. G. Lynch Jr., and R. G. Netemeyer. 2014. Financial literacy, financial education, and downstream financial behaviors. Management Science 60 (8): 1861-1883.

Fornero, E., and C. Monticone. 2011. Financial literacy and pension plan participation in Italy. Journal of Pension Economics and Finance 10 (4): 547–564.

Fort, M., F. Manaresi, and S. Trucchi. 2016. Adult financial literacy and households’ financial assets: the role of bank information policies. Economic Policy 31 (88): 743-782.

French, M. T., J. F. Homer, I. Popovici, and P. K. Robins. 2015. What you do in high school matters: High

28

school GPA, educational attainment, and labor market earnings as a young adult. Eastern Economic Journal 41: 370-386.

Grohmann, A., R. Kouwenberg, and L. Menkhoff. 2015. Childhood roots of financial literacy. Journal of Economic Psychology 51: 114-133.

Hastings, J. S., B. C. Madrian, and W. L. Skimmyhorn. 2013. Financial literacy, financial education, and economic outcomes. Annual Review of Economics 5 (1): 347-373.

Hong, H., J. D. Kubik, and J. C. Stein. 2004. Social interaction and stock-market participation. The Journal of Finance 59 (1): 137–163.

Hospido, L., E. Villanueva, and G. Zamarro. 2015. Finance for all: The impact of financial literacy training in compulsory secondary education in Spain. IZA Discussion Paper No. 8902.

Huston, S. J. 2012. Financial literacy and the cost of borrowing. International Journal of Consumer Studies 36: 566-572.

Jappelli, T., and M. Padula. 2013. Investment in financial literacy and saving decisions. Journal of Banking and Finance 37 (8): 2779–2792.

Johnson, and M. S. Sherraden. 2007. From financial literacy to financial capability among youth. The Journal of Sociology & Social Welfare 34 (3): 119-145.

Jorgensen, B. L., and J. Savla. 2010. Financial literacy of young adults: The importance of parental socialization. Family Relations 59 (4): 465–478.�

Klapper, L. F., A. Lusardi, and G. A. Panos. 2013. Financial literacy and its consequences: Evidence from Russia during the financial crisis. Journal of Banking and Finance 37 (10): 3904-3923.

Lührmann, M., M. Serra-Garcia, and J. Winter. 2015. Teaching teenagers in finance: Does it work? Journal of Banking and Finance 54 (May): 160-174.

Lusardi, A. (2015a). Financial literacy skills for the 21st century: Evidence from PISA. Journal of Consumer Affairs 49 (3): 639-659.

Lusardi, A. (2015b). Financial literacy: Do people know the ABC of finance? Public Understanding of Science 24 (3): 260-271.

Lusardi A., and O. S. Mitchell. 2007. Financial literacy and retirement preparedness: Evidence and implications for financial education. Business Economics 42 (January): 35–44.

Lusardi, A., and O. S. Mitchell. 2009. How ordinary consumers make complex economic decisions: Financial literacy and retirement readiness. National Bureau of Economic Research (NBER) Working Paper 15350.

Lusardi, A., and O. S. Mitchell. 2011. Financial literacy and retirement planning in the United States. Journal of Pension Economics and Finance 10 (4): 509–525.

Lusardi, A., and O. S. Mitchell. 2014. The economic importance of financial literacy. Journal of Economic

29

Literature 52 (1): 5-44.

Lusardi, A., O. S. Mitchell, and V. Curto. 2010. Financial literacy among the young. Journal of Consumer Affairs 44 (2): 358–380.

Lusardi, A., and P. Tufano. 2015. Debt literacy, financial experiences, and overindebtedness. Journal of Pension Economics and Finance 14 (4): 332-368.

Mahdavi, M., and N. Horton. 2014. Financial knowledge among educated women: Room for improvement. Journal of Consumer Affairs 48 (2): 403-471.

Mandell, L. 2005. The financial literacy of young African-American adults. Report prepared for Operation Hope.

Mandell, L. 2008. Financial education in high school. In A. Lusardi (Ed.), Overcoming the Saving Slump: How to Increase the Effectiveness of Financial Education and Saving Programs, The University of Chicago Press: 257-259.

OECD. 2014a. PISA 2012 Results: Students and Money: Financial Literacy Skills for the 21st Century (Volume VI), PISA, Paris, OECD Publishing.

OECD. 2014b. PISA 2012 Technical Report, Paris, OECD Publishing.

Sekita, S. 2011. Financial literacy and retirement planning in Japan. Journal of Pension Economics and Finance 10 (4): 637-656.

Shim, S., B. L. Barber, N. A. Card, J. J. Xiao, and J. Serido. 2010. Financial socialization of first-year college students: The roles of parents, work and education. Journal of Youth and Adolescence 39: 1457–1470.�

Van Campenhout, G. 2015. Revaluing the role of parents as financial socialization agents in youth financial literacy programs. Journal of Consumer Affairs 49 (1): 186-222.

van Rooij, M., A. Lusardi, and R. Alessie. 2011. Financial literacy and stock market participation. Journal of Financial Economics 101 (2): 449–472.

van Rooij, M., A. Lusardi, and R. Alessie. 2012. Financial literacy, retirement planning and household wealth. Economic Journal 122 (560): 449–478.

Walstad, W. B., K. Rebeck, and R. A. MacDonald. 2010. The effects of financial education on the financial knowledge of high school students. Journal of Consumer Affairs 44 (2): 336–357.

Xia, T., Z. Wang, and K. Li. 2014. Financial literacy, overconfidence and stock market participation. Social Indicators Research 119 (3): 1233-1245.

Xiao, J. J., S. Y. Ahn, J. Serido, and S. Shim. 2014. Earlier financial literacy and later financial behaviour of college students. International Journal of Consumer Studies 38 (6): 593-601.

Xu, L., and B. Zia. 2012. Financial literacy around the world. An overview of the evidence with practical suggestions for the way forward. World Bank Policy Research Working Paper 6107.

30

Tables and Figures Table 1: School outcomes as proxies for students’ perceived value of schooling

Variable Wording Coding

a. Time commitment to education

Truancy

Late for schoolIn the last two full weeks of school, how many times did you arrive late for school?

Skip whole dayIn the last two full weeks of school, how many times did you skip a whole school day?

Skip classes In the last two full weeks of school, how many times did you skip some classes?

Time spent on activities

Homework out of school Thinking about all school subjects: on average, how many hours do you spend each week on homework or other study set by your teachers?

Hours per week

b. Attitudes*

Attitudes towards school

Prepare for life School has contributed to prepare me for adult life

Valuable time School has not been a waste of time

Get a job Trying hard at school will help me get a good job

Prepare for college Trying hard at school will help me get into a good

Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database

* Variables have been recoded such that higher ordinal values are associated with more positive attitudes towards school.

1: Strongly disagree2: Disagree

3: Agree4: Strongly agree

0: None1: One or twice

2: Three or four times3: Five or more times

31

Table 2: Description of PISA 2012 proficiency levels on financial literacy scale

Proficiency level and lower cut score

Task descriptions

Level 5

625

Students can apply their understanding of a wide range of financial terms and concepts tocontexts that may only become relevant to their lives in the long term. They can analyzecomplex financial products and can take into account features of financial documents that aresignificant but unstated or not immediately evident, such as transaction costs. They can workwith a high level of accuracy and solve non-routine financial problems, and they can describethe potential outcomes of financial decisions, showing an understanding of the wider financiallandscape, such as income tax.

Level 4

550

Students can apply their understanding of less common financial concepts and terms tocontexts that will be relevant to them as they move towards adulthood, such as bank accountmanagement and compound interest in saving products. They can interpret and evaluate arange of detailed financial documents, such as bank statements, and explain the functions ofless commonly used financial products. They can make financial decisions taking into accountlonger-term consequences, such as understanding the overall cost implication of paying back aloan over a longer period, and they can solve routine problems in less common financialcontexts.

Level 3

475

Students can apply their understanding of commonly used financial concepts, terms andproducts to situations that are relevant to them. They begin to consider the consequences offinancial decisions and they can make simple financial plans in familiar contexts. They canmake straightforward interpretations of a range of financial documents and can apply a rangeof basic numerical operations, including calculating percentages. They can choose thenumerical operations needed to solve routine problems in relatively common financial literacycontexts, such as budget calculations.

Level 2

400

Students begin to apply their knowledge of common financial products and commonly usedfinancial terms and concepts. They can use given information to make financial decisions incontexts that are immediately relevant to them. They can recognize the value of a simplebudget and can interpret prominent features of everyday financial documents. They can applysingle basic numerical operations, including division, to answer financial questions. They showan understanding of the relationships between different financial elements, such as theamount of use and the costs incurred.

Level 1

326

Students can identify common financial products and terms and interpret information relatingto basic financial concepts. They can recognize the difference between needs and wants andcan make simple decisions on everyday spending. They can recognize the purpose of everydayfinancial documents such as an invoice and apply single and basic numerical operations(addition, subtraction or multiplication) in financial contexts that they are likely to haveexperienced personally.

Note: To reach a particular proficiency level, a student must correctly answer a majority of items at that level. Students were classified into financial literacy levels

according to their scores. Cut scores in the exhibit are rounded. Scores are reported on a scale from 0 to 1,000.

Source: Organization for Economic Cooperation and Development (OECD), Program for International Student Assessment (PISA), 2012.

32

Table 3: Regional-level instruments for financial literacy

Region

Share of newspapers with finance content

over total volume in 2012

Growth of ATM branches

between 2010 and 2012

Number of bankruptcies in

2012

Abruzzo 0.0119 -0.020 281

Basilicata 0.015 0.008 55

Calabria 0.0162 -0.025 285

Campania 0.0117 -0.005 950

Emilia-Romagna 0.0139 -0.021 950

Friuli 0.0118 -0.015 253

Lazio 0.0163 -0.009 1,252

Liguria 0.0328 -0.013 257

Lombardia 0.0215 -0.015 2,613

Marche 0.0142 -0.033 417

Molise 0.0115 0.000 42

Piemonte 0.0128 -0.006 881

Puglia 0.0116 -0.008 497

Sardegna 0.0438 0.001 242

Sicilia 0.0106 -0.037 610

Toscana 0.0121 -0.002 786

Trentino-Alto Adige 0.0171 -0.003 137

Umbria 0.0149 0.019 210

Valle d'Aosta 0.0166 0.021 17

Veneto 0.0130 -0.010 1,021

Average 0.0164 -0.009 589

Note : Author's calculations from raw data, available upon request

Sources: Accertamenti diffusione stampa (ADS); Bank of Italy; Italian Ministry of Justice

33

Table 4: Descriptive statistics for dependent variables of interest

Outcomes N Mean (SD) Min Max

a. Time commitment to education

Truancy

Late for school 7,029 0.50 (0.77) 0 3

Skip whole day 7,026 0.59 (0.69) 0 3

Skip classes 7,019 0.43 (0.66) 0 3

Time spent on activities*

Homework out of school 4,547 8.75 (7.09) 0 30

b. Attitudes

Attitudes towards school*

Prepare for life 4,612 2.76 (0.81) 1 4

Valuable time 4,620 3.17 (0.75) 1 4

Get a job 4,607 3.32 (0.69) 1 4

Prepare for college 4,596 3.24 (0.73) 1 4

Note : Stratified two-stage sample design and replicate weights accounted for

* Time spent on activities not asked in questionnaire B; attitudes not asked in questionnaire A

Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database

34

Table 5: Descriptive statistics on sample of schools and students, by region

Region# of

schools# of

students

Vocational orientation

(%)

Private (%)

# students per school interviewed

Average school size

Average class size

Student-teacher ratio

North-West

Liguria 58 331 43.2% 7.8% 6.5 654.0 23.1 10.0(1.85) (344.57) (3.68) (3.50)

Lombardia 54 347 49.6% 8.5% 6.9 879.9 28.3 10.7(1.67) (441.78) (11.84) (3.51)

Piemonte 51 339 57.0% 9.9% 7.1 664.4 26.6 9.7(1.29) (382.53) (10.91) (2.42)

Valle d'Aosta 30 178 46.0% 13.0% 9.1 360.4 22.4 7.7(4.35) (302.61) (10.39) (3.04)

North-East

Emilia-Romagna 54 331 59.4% 4.8% 6.8 741.5 26.2 10.0(1.67) (404.48) (9.46) (3.39)

Friuli 53 330 49.6% 8.0% 6.8 786.1 22.8 10.2(1.63) (555.99) (7.59) (3.36)

Trentino-Alto Adige 136 794 59.9% 13.1% 6.9 577.0 24.9 8.2(1.47) (376.76) (10.83) (3.85)

Veneto 70 455 55.0% 12.9% 7.1 786.2 25.1 11.6(1.61) (442.38) (7.57) (3.91)

Center

Lazio 54 340 41.6% 4.2% 6.7 764.5 24.4 10.5(1.54) (390.31) (7.18) (3.26)

Marche 53 336 52.6% 0.0% 7.0 738.0 24.6 10.0(1.76) (309.82) (5.88) (2.45)

Toscana 54 320 55.4% 2.8% 6.5 708.0 23.4 9.5(1.54) (425.96) (3.49) (2.93)

Umbria 53 325 45.1% 0.7% 6.7 654.5 24.3 10.0(1.69) (315.55) (6.64) (2.81)

South

Abruzzo 53 336 42.0% 2.6% 6.8 718.8 24.8 9.3(1.47) (406.91) (7.24) (3.26)

Basilicata 56 351 52.7% 0.4% 6.7 496.9 22.6 9.7(1.59) (328.85) (5.19) (3.34)

Calabria 58 354 46.1% 0.7% 6.6 673.1 24.0 8.9(1.92) (367.04) (8.53) (3.54)

Campania 53 347 44.3% 4.0% 7.0 852.5 26.1 10.8(1.51) (362.69) (8.36) (3.20)

Molise 48 254 46.6% 0.0% 6.8 495.3 27.6 8.83(1.57) (251.86) (11.72) (2.7)

Puglia 56 365 47.4% 0.2% 7.1 761.7 28.8 11.4(1.34) (363.18) (10.87) (3.96)

Islands

Sardegna 58 307 45.7% 1.1% 6.0 560.1 23.1 9.7(1.88) (333.59) (7.48) (2.76)

Sicilia 56 328 42.5% 0.5% 6.4 793.6 25.1 9.9(1.6) (472.22) (9.22) (2.56)

Total 1,158 7,068 49.1% 5.2% 6.8 762.6 25.8 10.4(1.62) (413.64) (9.23) (3.38)

Note : Stratified two-stage sample design and replicate weights accounted for. Standard deviation in parentheses.

Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database

35

Figure 1: Descriptive statistics on financial literacy competence, by region

Source: Author’s computations from the OECD PISA 2012 Financial Literacy (FL) database

415.5

428.6

439.3445.6 446.2 449.0 452.8

460.0 462.5 466.3 467.7 471.0 474.1 474.1 476.1480.8 481.1

491.5499.0 501.5 501.5

300

350

400

450

500

Mea

n fin

ancia

l lite

racy

sco

res

36

Figure 2: Share of students at each proficiency level on the financial literacy scale, by region

Source: Author’s computations from the OECD PISA 2012 Financial Literacy (FL) database

0 10 20 30 40 50 60 70 80 90 100

Friuli

Veneto

Trentino Alto-Adige

Lombardia

Piemonte

Umbria

Emilia-Romagna

Marche

Valle d'Aosta

Toscana

Liguria

Average

Puglia

Basilicata

Lazio

Abruzzo

Molise

Sardegna

Campania

Sicilia

Calabria

Percentage of Students at Each Level of Proficiency in Financial Literacy

Level 1 or below

Level 2

Level 3

Level 4

Level 5 and above

37

Table 6: OLS estimates of financial literacy and students’ perceived value of schooling (panels a and b)

correct

(1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4)Financial literacy -0.074*** -0.064*** -0.066*** -0.047*** -0.081*** -0.062*** -0.055*** -0.044*** -0.048*** -0.044*** -0.048*** -0.036*** 2.141*** 2.109*** 1.862*** 1.304***

(0.013) (0.014) (0.015) (0.016) (0.010) (0.011) (0.012) (0.014) (0.009) (0.010) (0.010) (0.011) (0.128) (0.137) (0.147) (0.168)Child characteristics

Female (Ref. Male) -0.067** -0.073** -0.052* -0.037* -0.052** -0.027 -0.103*** -0.107*** -0.091*** 3.075*** 3.256*** 2.591***(0.027) (0.029) (0.030) (0.022) (0.021) (0.023) (0.025) (0.025) (0.028) (0.268) (0.274) (0.294)

Age -0.015 -0.004 -0.013 0.049 0.038 0.040 -0.004 -0.003 -0.015 -0.499 -0.276 -0.413(0.051) (0.049) (0.047) (0.037) (0.038) (0.036) (0.031) (0.030) (0.032) (0.535) (0.514) (0.533)

Grade -0.067** -0.055* -0.067** -0.117*** -0.089*** -0.092*** -0.034 -0.032 -0.038 0.650** 0.318 0.535**(0.030) (0.031) (0.033) (0.032) (0.032) (0.035) (0.030) (0.031) (0.034) (0.254) (0.259) (0.266)

Household characteristicsESCS Index 0.008 0.018 -0.033*** -0.007 0.025* 0.029** 1.209*** 0.757***

(0.014) (0.016) (0.012) (0.015) (0.013) (0.014) (0.155) (0.181)Two-parent family (Ref. Other) -0.151*** -0.141** -0.146*** -0.162*** -0.058* -0.066 0.866** 0.365

(0.056) (0.061) (0.041) (0.044) (0.035) (0.042) (0.418) (0.406)School characteristics

Private school (Ref. Public) -0.120* -0.180*** -0.056 -0.477(0.066) (0.061) (0.061) (0.783)

Vocational (Ref. General) 0.066** 0.112*** 0.056* -3.144***(0.032) (0.032) (0.030) (0.306)

Class size -0.001 -0.002* 0.000 -0.003(0.002) (0.001) (0.001) (0.012)

Total school enrolment -0.000** -0.000* -0.000 0.000(0.000) (0.000) (0.000) (0.000)

Student-Teacher ratio -0.000 -0.001 0.005 0.181***(0.005) (0.004) (0.004) (0.046)

Constant 0.869*** 1.067 1.068 1.089 1.072*** 0.203 0.469 0.404 0.665*** 0.757 0.827* 0.892* -2.749*** 3.745 0.955 7.105(0.092) (0.821) (0.798) (0.753) (0.078) (0.615) (0.634) (0.587) (0.072) (0.489) (0.484) (0.527) (0.923) (8.520) (8.174) (8.542)

Observations 7,029 7,029 6,756 5,957 7,026 7,026 6,754 5,956 7,019 7,019 6,750 5,952 4,547 4,547 4,374 3,854R-squared 0.031 0.036 0.039 0.042 0.041 0.049 0.054 0.071 0.011 0.018 0.02 0.021 0.13 0.18 0.208 0.27Note : Stratified two-stage sample design and replicate weights accounted for. Robust standard errors clustered at the school level (in parentheses). Regional dummies included. *** p<0.01, ** p<0.05, * p<0.1Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database.

a. Time commitment to education Late for school Skip whole day Skip classesTruancy Time spent on activities

Homework out of school

38

(1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4)Financial literacy 0.102*** 0.114*** 0.112*** 0.105*** 0.144*** 0.146*** 0.133*** 0.119*** -0.038*** -0.039*** -0.039*** -0.027* 0.100*** 0.081*** 0.071*** 0.040**

(0.018) (0.021) (0.021) (0.022) (0.015) (0.016) (0.018) (0.019) (0.012) (0.014) (0.015) (0.016) (0.015) (0.016) (0.015) (0.017)Child characteristics

Female (Ref. Male) 0.208*** 0.193*** 0.176*** 0.324*** 0.313*** 0.286*** 0.069** 0.074** 0.104*** 0.194*** 0.194*** 0.142***(0.039) (0.040) (0.042) (0.031) (0.031) (0.034) (0.031) (0.032) (0.035) (0.029) (0.029) (0.030)

Age -0.057 -0.056 -0.080 -0.099** -0.083* -0.102** 0.082* 0.079 0.080 0.071 0.088* 0.091(0.058) (0.055) (0.058) (0.046) (0.044) (0.048) (0.049) (0.048) (0.054) (0.053) (0.051) (0.058)

Grade -0.040 -0.037 -0.017 0.035 0.038 0.071* 0.009 -0.004 -0.057 0.121*** 0.096** 0.082*(0.040) (0.040) (0.040) (0.039) (0.041) (0.040) (0.044) (0.044) (0.035) (0.045) (0.044) (0.044)

Household characteristicsESCS Index -0.022 -0.040** 0.042** 0.014 0.009 0.009 0.057*** 0.018

(0.020) (0.020) (0.017) (0.019) (0.015) (0.016) (0.017) (0.017)Two-parent family (Ref. Other) 0.042 0.045 0.026 0.045 0.084 0.029 0.098* 0.048

(0.065) (0.065) (0.055) (0.054) (0.052) (0.057) (0.058) (0.060)School characteristics

Private school (Ref. Public) 0.240** -0.136 0.080 -0.076(0.105) (0.114) (0.079) (0.071)

Vocational (Ref. General) -0.026 -0.069* 0.095*** -0.215***(0.043) (0.038) (0.034) (0.033)

Class size -0.001 -0.001 -0.003 -0.000(0.002) (0.002) (0.002) (0.002)

Total school enrolment 0.000 0.000 0.000 0.000(0.000) (0.000) (0.000) (0.000)

Student-Teacher ratio -0.000 0.006 0.001 0.004(0.005) (0.004) (0.005) (0.005)

Constant 2.251*** 2.971*** 2.927*** 3.346*** 2.316*** 3.720*** 3.537*** 3.865*** 3.593*** 2.278*** 2.239*** 2.164** 2.698*** 1.615* 1.318 1.593*(0.133) (0.908) (0.859) (0.901) (0.124) (0.709) (0.684) (0.749) (0.077) (0.768) (0.769) (0.877) (0.094) (0.826) (0.815) (0.949)

Observations 4,612 4,612 4,443 3,928 4,620 4,620 4,450 3,932 4,607 4,607 4,438 3,923 4,596 4,596 4,427 3,915R-squared 0.024 0.041 0.037 0.041 0.052 0.1 0.098 0.108 0.016 0.02 0.02 0.026 0.036 0.063 0.069 0.083Note : Stratified two-stage sample design and replicate weights accounted for. Robust standard errors clustered at the school level (in parentheses). Regional dummies included. *** p<0.01, ** p<0.05, * p<0.1

Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database.

Prepare for college Attitudes towards school

b. Attitudes Get a jobPrepare for life Valuable time

39

Table 7: OLS and IV estimates of financial literacy on students’ perceived value of schooling (full specification)

(1) (2) (3) (4)OLS Regional IVs Individual IV Combined IVs

a. Time commitment to education

Truancy

Late for school -0.047*** -0.264** -0.181** -0.220***(0.016) (0.103) (0.091) (0.074)

Skip whole day -0.044*** -0.314*** -0.164** -0.245***(0.014) (0.088) (0.081) (0.060)

Skip classes -0.036*** -0.128* 0.013 -0.067(0.011) (0.074) (0.076) (0.050)

Time spent on activities

Homework out of school 1.304*** -0.340 2.201** 1.325**(0.168) (0.882) (1.081) (0.589)

b. Attitudes

Attitudes towards school

Prepare for life 0.105*** 0.041 -0.014 0.006(0.022) (0.131) (0.132) (0.097)

Valuable time 0.119*** -0.109 -0.040 -0.079(0.019) (0.119) (0.110) (0.089)

Get a job -0.027* -0.035 -0.174 -0.097(0.016) (0.103) (0.113) (0.076)

Prepare for college 0.040** -0.141 -0.017 -0.081(0.017) (0.104) (0.115) (0.082)

Note : Stratified two-stage sample design and replicate weights accounted for. Robust standard errors clustered at the school level (in parentheses).

*** p<0.01, ** p<0.05, * p<0.1

Sources: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database, combined with data from Accertamenti Diffusione

Stampa (ADS), Bank of Italy, and Italian Ministry of Justice.

Financial LiteracyDependent variable

40

Appendix A Table A1: Students’ performance in financial literacy across 18 economies

Mean (SD)

Australia 526.05 (101.24)

Belgium 541.10 (97.46)

Colombia 378.66 (105.64)

Czech Republic 513.19 (88.32)

Spain 484.25 (85.18)

Estonia 529.06 (79.01)

France 486.26 (105.55)

Croatia 480.30 (85.18)

Israel 476.46 (115.41)

Italy 466.33 (87.19)

Latvia 500.60 (77.96)

New Zealand 519.98 (118.04)

Poland 510.13 (81.91)

Shanghai-China 603.38 (83.49)

Russian Federation 486.27 (87.82)

Slovak Republic 470.45 (104.98)

Slovenia 484.10 (89.13)

United States 491.60 (99.36)

OECD average 500 (100)Note : Stratified two-stage sample design and replicates accounted for

Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database

Financial LiteracyCountry

41

Table A2: Correlation matrices for continuous variables

Truancy (1) (2) (3)

(1) Late for school 1

(2) Skip whole day 0.2296 1

(3) Skip classes 0.2559 0.2607 1

Attitudes towards school (1) (2) (3) (4)

(1) Prepare for life 1

(2) Valuable time 0.4443 1

(3) Get a job 0.1235 0.2248 1

(4) Prepare for college 0.1592 0.2796 0.479 1

Regional-level instruments (1) (2) (3)

(1) Newspapers 1

(2) Growth of ATM -0.0668 1

(3) Bankruptcies 0.2304 -0.2482 1Sources: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database, combined with data from

Accertamenti Diffusione Stampa (ADS), Bank of Italy, and Italian Ministry of Justice.

42

Table A3: F-tests for first-stage IV regression specifications

First stage (F-test)Regional-level

Share of newspapers with finance-related content 3.79Growth of ATM branches 2010-2012 8.14Number of bankruptcies in 2012 9.39Combined 12.73

Individual-level

Share of schoolmates with parents working in money-related occupation 13.05

Regional and individual-level combined 25.52Sources: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database, combined

with data from Accertamenti diffusione stampa (ADS), Bank of Italy, and Italian Ministry of Justice

43

Table A4: Missing data analysis

Time spent on activitiesLate for school Skip whole day Skip classes Homework out of school Prepare for life Valuable time Get a job Prepare for college

Age 0.049 -0.028 0.034 -0.001 0.000 0.004 0.004 0.001(0.074) (0.068) (0.058) (0.010) (0.012) (0.012) (0.012) (0.012)

Female -0.015 0.009 0.077 -0.027 0.023 0.023 0.019 0.019(0.095) (0.101) (0.086) (0.017) (0.016) (0.016) (0.016) (0.016)

Grade 0.006 -0.186 -0.063 0.005 0.002 0.003 0.002 0.001(0.125) (0.174) (0.142) (0.018) (0.017) (0.017) (0.017) (0.017)

ESCS Index -0.043 -0.190 -0.230 -0.013 0.029 0.031 0.027 0.023(0.223) (0.223) (0.174) (0.036) (0.035) (0.036) (0.035) (0.035)

Financial Literacy -61.368*** -34.602** -56.626*** -1.334 -5.398* -4.854 -5.496* -6.047*(13.522) (17.321) (17.585) (3.376) (3.173) (3.160) (3.099) (3.087)

% missing 0.61% 0.79% 0.85% 35.57% 34.80% 34.66% 34.82% 35.05%Note: Robust standard errors clustered at the school level (in parentheses); *** p<0.01, ** p<0.05, * p<0.1This table reports coefficients from a regression of the variables listed on the left on a dummy for whether the dependent variable is missing. This is analogous to a ttest accounting for the sampling design.Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database

a. Time commitment to education b. AttitudesTruancy Attitudes towards school

44

Table A5: OLS and IV estimates of financial literacy on students’ perceived value of schooling, all specifications

(1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4) (1) (2) (3) (4)

a. Time commitment to education

Truancy

Late for school -0.074*** -0.064*** -0.066*** -0.047*** -0.276*** -0.268*** -0.292*** -0.264** -0.148** -0.136** -0.175** -0.181** -0.184*** -0.188*** -0.222*** -0.220***(0.013) (0.014) (0.015) (0.016) (0.085) (0.076) (0.086) (0.103) (0.059) (0.067) (0.076) (0.091) (0.049) (0.051) (0.057) (0.074)

Skip whole day -0.081*** -0.062*** -0.055*** -0.044*** -0.286*** -0.276*** -0.297*** -0.314*** -0.184*** -0.170*** -0.159*** -0.164** -0.215*** -0.214*** -0.218*** -0.245***(0.010) (0.011) (0.012) (0.014) (0.081) (0.073) (0.080) (0.088) (0.044) -0.049 (0.059) (0.081) (0.039) (0.042) (0.048) (0.060)

Skip classes -0.048*** -0.044*** -0.048*** -0.036*** -0.048 -0.064 -0.081 -0.128* -0.036 -0.008 -0.007 0.013 -0.039 -0.030 -0.040 -0.067(0.009) (0.010) (0.010) (0.011) (0.067) (0.059) (0.066) (0.074) (0.042) (0.047) (0.056) (0.076) (0.035) (0.036) (0.041) (0.050)

Time spent on activities

Homework out of school 2.141*** 2.109*** 1.862*** 1.304*** 0.365 0.544 0.083 -0.340 3.630*** 3.121*** 2.644*** 2.201** 2.621*** 2.019*** 1.487*** 1.325**(0.128) (0.137) (0.147) (0.168) (0.924) (0.800) (0.831) (0.882) (0.679) (0.706) (0.810) (1.081) (0.521) (0.512) (0.565) (0.589)

b. Attitudes

Attitudes towards school

Prepare for life 0.102*** 0.114*** 0.112*** 0.105*** 0.060 0.057 0.092 0.041 0.050 0.015 0.010 -0.014 0.050 0.026 0.036 0.006(0.018) (0.021) (0.021) (0.022) (0.125) (0.109) (0.118) (0.131) (0.069) (0.081) (0.094) (0.132) (0.063) (0.068) (0.077) (0.097)

Waste of time 0.144*** 0.146*** 0.133*** 0.119*** -0.152 -0.121 -0.141 -0.109 0.148** 0.077 0.026 -0.040 0.068 0.002 -0.042 -0.079(0.015) (0.016) (0.018) (0.019) (0.126) (0.102) (0.111) (0.119) (0.065) (0.069) (0.079) (0.110) (0.058) (0.061) (0.069) (0.089)

Get a job -0.038*** -0.039*** -0.039*** -0.027* -0.060 -0.068 -0.057 -0.035 -0.094 -0.123* -0.167** -0.174 -0.080 -0.096* -0.119* -0.097(0.012) (0.014) (0.015) (0.016) (0.101) (0.093) (0.099) (0.103) (0.063) (0.069) (0.081) (0.113) (0.052) (0.054) (0.061) (0.076)

Prepare for college 0.100*** 0.081*** 0.071*** 0.040** -0.132 -0.094 -0.131 -0.141 0.155** 0.091 0.053 -0.017 0.073 0.018 -0.022 -0.081(0.015) (0.016) (0.015) (0.017) (0.118) (0.099) (0.106) (0.104) (0.066) (0.072) (0.081) (0.115) (0.058) (0.061) (0.068) (0.082)

Note : Stratified two-stage sample design and replicate weights accounted for. Robust standard errors clustered at the school level (in parentheses). *** p<0.01, ** p<0.05, * p<0.1

Source: Author's computations from the OECD PISA 2012 Financial Literacy (FL) database, combined with data from Accertamenti Diffusione Stampa (ADS), Bank of Italy, and Italian Ministry of Justice.

Combined IVFinancial Literacy

Dependent variable OLS Regional-level IVs Individual-level IV